Re: [RFC PATCH v2 20/21] x86: Add support for CONFIG_CFI_CLANG

[Sami Tolvanen]

On Mon, May 16, 2022 at 2:54 AM Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:
>
> On Fri, May 13, 2022 at 01:21:58PM -0700, Sami Tolvanen wrote:
> > With CONFIG_CFI_CLANG, the compiler injects a type preamble
> > immediately before each function and a check to validate the target
> > function type before indirect calls:
> >
> >   ; type preamble
> >   __cfi_function:
> >     int3
> >     int3
> >     mov <id>, %eax
> >     int3
> >     int3
> >   function:
> >     ...
>
> When I enable CFI_CLANG and X86_KERNEL_IBT I get:
>
> 0000000000000c80 <__cfi_io_schedule_timeout>:
> c80:   cc                      int3
> c81:   cc                      int3
> c82:   b8 b5 b1 39 b3          mov    $0xb339b1b5,%eax
> c87:   cc                      int3
> c88:   cc                      int3
>
> 0000000000000c89 <io_schedule_timeout>:
> c89:   f3 0f 1e fa             endbr64
>
>
> That seems unfortunate. Would it be possible to get an additional
> compiler option to suppress the endbr for all symbols that get a __cfi_
> preaamble?

What's the concern with the endbr? Dropping it would currently break
the CFI+IBT combination on newer hardware, no?

> Also, perhaps s/CFI_CLANG/KERNEL_CFI/ or somesuch, so that GCC might
> also implement this same scheme (in time)?

I'm fine with renaming the config.

> >   ; indirect call check
> >     cmpl    <id>, -6(%r11)
> >     je      .Ltmp1
> >     ud2
> >   .Ltmp1:
> >     call    __x86_indirect_thunk_r11
>
> The first one I try and find looks like:
>
> 26:       41 81 7b fa a6 96 9e 38         cmpl   $0x389e96a6,-0x6(%r11)
> 2e:       74 02                   je     32 <__traceiter_sched_kthread_stop+0x29>
> 30:       0f 0b                   ud2
> 32:       4c 89 f6                mov    %r14,%rsi
> 35:       e8 00 00 00 00          call   3a <__traceiter_sched_kthread_stop+0x31> 36: R_X86_64_PLT32      __x86_indirect_thunk_r11-0x4
>
> This must not be. If I'm to rewrite that lot to:
>
>   movl  $\hash, %r10d
>   sub   $9, %r11
>   call  *%r11
>   .nop  4
>
> Then there must not be spurious instruction in between the ud2 and the
> indirect call/retpoline thing.

With the current compiler patch, LLVM sets up function arguments after
the CFI check. if it's a problem, we can look into changing that.

Sami